Physical and Physicochemical Classification of Parboiled Rice Using VNIR-SWIR Spectroscopy and Machine Learning

doi:10.1016/j.rsci.2025.08.007

Abstract

Abstract:

The classification of parboiled rice into types can be optimized through the use of machine learning (ML) algorithms, resulting in greater speed and accuracy in data processing. The objectives of this study were: (i) to investigate the spectral behavior of different types of parboiled rice (Types 1-5 and Off-type); (ii) to identify the most effective ML algorithm for classifying parboiled rice types; (iii) to determine the best kernel configuration and preprocessing methods for spectral data; and (iv) to recommend a protocol for implementing this technique in the rice storage industry. Samples were selected based on the maximum defect limits tolerated for each type, according to the Technical Rice Regulation. Spectral data were acquired using a spectroradiometer in the range of 350-2500 nm and subsequently processed with different methods, including baseline correction, standard normal variate, multiplicative scattering correction, combinations of these techniques with Savitzky-Golay smoothing, and the application of the first derivative of Savitzky-Golay smoothing. The data were analyzed using six different ML algorithms: Artificial Neural Network, Decision Tree, Logistic Regression, REPTree, Random Forest, and Support Vector Machine. Rice types were treated as output variables, while spectral features served as input variables. Logistic Regression and Support Vector Machine algorithms showed the best classification performance, with accuracy rates above 97%, F-scores around 0.98, and Kappa values exceeding 0.97. Spectral preprocessing did not yield substantial improvements and incurred high computational costs; therefore, using raw data was a viable and efficient alternative. For practical implementation in the rice storage industry, we recommend acquiring a VNIR-SWIR (visible near-infrared and shortwave infrared) hyperspectral sensor (350-2500 nm) and developing a classification model based on the Support Vector Machine algorithm with a linear kernel trained on representative local samples. Additionally, we recommend implementing an automated real-time classification system, a representative sample collection protocol, and detailed reporting for inventory and logistics optimization.

Key words: Oryza sativa L., artificial intelligence, supervised classification, support vector machine, logistic regression

Nairiane dos Santos Bilhalva, Paulo Carteri Coradi, Rosana Santos de Moraes, Dthenifer Cordeiro Santana, Larissa Ribeiro Teodoro, Paulo Eduardo Teodoro, Marisa Menezes Leal. Physical and Physicochemical Classification of Parboiled Rice Using VNIR-SWIR Spectroscopy and Machine Learning[J]. Rice Science, 2025, 32(6): 857-867.

Figures/Tables 7

Fig. 1. Hyperspectral curve from Types 1-5 and Off-type of parboiled rice. A, Visible region (350-750 nm). B, Near-infrared region (750-1300 nm). C, Shortwave-infrared region (1300-1500 nm).

Fig. 2. Reflectance spectra of rice samples after preprocessing steps. A, Raw reflectance spectra; B, Spectra after baseline correction; C, Standard normal variate (SNV) transformation; D, Multiplicative scatter correction (MSC); E, Combined preprocessing (SNV and MSC) using Savitzky-Golay smoothing; F, First derivative spectra after Savitzky-Golay smoothing. These transformations progressively reduce noise and scattering effects, enhancing spectral differences among rice types.

Fig. 3. Principal component analysis (PCA) for parboiled rice samples. A, Score plots for the first three principal components of all parboiled rice sample; B, PCA loading plot of spectral data without preprocessing.

Fig. 4. Boxplots showing distribution of performance metrics across different machine learning algorithms. A, Correlation coefficient (CC). B, F-score. C, Kappa between machine learning algorithms. DT, Decision Tree; J48, J48 Decision Tree algorithm; RF, Random Forest; ANN, Artificial Neural Network; LR, Logistic Regression; SVM, Support Vector Machine. Different lowercase letters above bars represent significant differences at the 0.05 level by the Scott-Knott test.

Fig. 5. Confusion matrices showing classification accuracy for parboiled rice types using Support Vector Machine (SVM, A) and Logistic Regression algorithms (B). Data show the percentage (%) of correct classifications obtained for each type.

Table 1. Correct classification rates among kernel functions and preprocessing methods obtained for training set (450 samples) and internal validation using Support Vector Machine algorithm.

Preprocessing	Linear kernel		Radial kernel		Polynomial kernel
Preprocessing	Training	Internal validation	Training	Internal validation	Training	Internal validation
No preprocessing	1.000	0.971	1.000	0.895	1.000	0.964
Baseline offset correction (BL)	1.000	0.980	1.000	0.915	0.997	0.973
Standard normal variate (SNV)	1.000	0.982	1.000	0.933	1.000	0.982
Multiplicative scatter correction	1.000	0.977	1.000	0.931	1.000	0.975
SNV + BL + Savitzky-Golay smoothing	1.000	0.975	1.000	0.928	1.000	0.977
First derivative of Savitzky-Golay	1.000	0.993	1.000	0.960	1.000	0.991

Table 2. Correct classification rate among kernel functions and preprocessing methods obtained for external validation set (149 samples) using Support Vector Machine algorithm.

Preprocessing	Kernel function
Preprocessing	Linear	Radial	Polynomial
No preprocessing	0.993	0.959	0.986
Baseline offset correction (BL)	0.986	0.953	0.986
Standard normal variate (SNV)	0.986	0.973	0.979
Multiplicative scatter correction	0.986	0.979	0.986
SNV + BL + Savitzky-Golay smoothing	0.986	0.986	0.979
First derivative of Savitzky-Golay	0.228	0.174	0.174

References 39

[1]	Al Snousy M B, El-Deeb H M, Badran K, et al. 2011. Suite of decision tree-based classification algorithms on cancer gene expression data. Egypt Inform J, 12(2): 73-82.
[2]	Aznan A, Viejo C G, Pang A, et al. 2023. Review of technology advances to assess rice quality traits and consumer perception. Food Res Int, 172: 113105.
[3]	Bagchi T B, Das B, Kumar A, et al. 2023. Impact of cooking, parboiling and fermentation on nutritional components, predicted glycemic index and pasting properties of rice. J Cereal Sci, 114: 103763.
[4]	Belgiu M, Drăguţ L. 2016. Random forest in remote sensing: A review of applications and future directions. ISPRS J Photogramm Remote Sens, 114: 24-31.
[5]	Bouckaert R R, Frank E, Hall M A, et al. 2010. WEKA: Experiences with a Java open-source project. J Mach Learn Res, 11: 2533-2541.
[6]	Cañizares L, Meza S, Peres B, et al. 2024. Functional foods from black rice (Oryza sativa L.): An overview of the influence of drying, storage, and processing on bioactive molecules and health-promoting effects. Foods, 13(7): 1088.
[7]	Çifci A, Kırbaş İ. 2025. Fusion of machine learning and explainable AI for enhanced rice classification: A case study on Cammeo and Osmancik species. Eur Food Res Technol, 251(1): 69-86.
[8]	Coradi P C, Müller A, da Silva André G, et al. 2021. Effects of cultivars and fertilization levels on the quality of brown and polished rice. Cereal Chem, 98(6): 1238-1249.
[9]	de Gregori G S, Loureiro E S, Pessoa L G A, et al. 2023. Machine learning in the hyperspectral classification of Glycaspis brimblecombei (Hemiptera: Psyllidae) attack severity in eucalyptus. Remote Sens, 15(24): 5657.
[10]	de Moraes R S, Coradi P C, Nunes M T, et al. 2023. Thick layer drying and storage of rice grain cultivars in silo-dryer-aerator: Quality evaluation at low drying temperature. Heliyon, 9(7): e17962.
[11]	Deepika S, Sutar P P. 2024. Spectral selective infrared heating of food components based on optical characteristics and penetration depth: A critical review. Crit Rev Food Sci Nutr, 64(29): 10749-10771.
[12]	Egmont-Petersen M, de Ridder D, Handels H. 2002. Image processing with neural networks: A review. Pattern Recognit, 35(10): 2279-2301.
[13]	Elmas Ş N K, Arslan F N, Akin G, et al. 2019. Synchronous fluorescence spectroscopy combined with chemometrics for rapid assessment of cold-pressed grape seed oil adulteration: Qualitative and quantitative study. Talanta, 196: 22-31.
[14]	Fattahi S H, Kazemi A, Khojastehnazhand M, et al. 2024. The classification of Iranian wheat flour varieties using FT-MIR spectroscopy and chemometrics methods. Expert Syst Appl, 239: 122175.
[15]	Ferreira D F. 2011. Sisvar: A computer statistical analysis system. Ciênc Agrotec, 35(6): 1039-1042.
[16]	Hao Y, Geng P, Wu W H, et al. 2019. Identification of rice varieties and transgenic characteristics based on near-infrared diffuse reflectance spectroscopy and chemometrics. Molecules, 24(24): 4568.
[17]	Hosseini S, Sardo S R. 2021. Data mining tool: A case study for network intrusion detection. Multimed Tools Appl, 80: 4999-5019.
[18]	Hu N Y, Li W, Du C H, et al. 2021. Predicting micronutrients of wheat using hyperspectral imaging. Food Chem, 343: 128473.
[19]	Iqbal S, Zhang P, Wu P, et al. 2021. Evolutions of rheology, microstructure and digestibility of parboiled rice during simulated semi-dynamic gastrointestinal digestion. LWT, 148: 111700.
[20]	Jin B, Zhang C, Jia L, et al. 2022. Identification of rice seed varieties based on near-infrared hyperspectral imaging technology combined with deep learning. ACS Omega, 7(6): 4735-4749.
[21]	Kumar A, Lal M K, Nayak S, et al. 2022. Effect of parboiling on starch digestibility and mineral bioavailability in rice (Oryza sativa L.). LWT, 156: 113026.
[22]	Leal A F, Coradi P C, dos Santos Moraes R, et al. 2023. Tempering time and air temperature relationships for real-scale paddy drying and their effects on the physical, physicochemical, and morphological qualities of polished rice. Dry Technol, 41(11): 1878-1892.
[23]	Lutz É, Coradi P C. 2022. Applications of new technologies for monitoring and predicting grains quality stored: Sensors, internet of things, and artificial intelligence. Measurement, 188: 110609.
[24]	Maldaner V, Coradi P C, Nunes M T, et al. 2021. Effects of intermittent drying on physicochemical and morphological quality of rice and endosperm of milled brown rice. LWT, 152: 112334.
[25]	Miao X X, Miao Y, Tao S H, et al. 2021. Classification of rice based on storage time by using near infrared spectroscopy and chemometric methods. Microchem J, 171: 106841.
[26]	Müller A, Nunes M T, Maldaner V, et al. 2022. Rice drying, storage and processing: Effects of post-harvest operations on grain quality. Rice Sci, 29(1): 16-30.
[27]	Nalepa J, Kawulok M. 2019. Selecting training sets for support vector machines: A review. Artif Intell Ver, 52(2): 857-900.
[28]	Nikzadfar M, Rashvand M, Zhang H W, et al. 2024. Hyperspectral imaging aiding artificial intelligence: A reliable approach for food qualification and safety. Appl Sci: Basel, 14(21): 9821.
[29]	Patel D, Bhise S, Kapdi S, et al. 2024. Non-destructive hyperspectral imaging technology to assess the quality and safety of food: A review. Food Prod Process Nutr, 6(1): 69.
[30]	Quinlan J R. 1993. C4.5: Programs for Machine Learning. CA, USA: Morgan Kauffmann Publishers Inc: 302.
[31]	Rajvanshi N, Chowdhary K R. 2017. Comparison of SVM and naïve Bayes text classification algorithms using WEKA. Int J Eng Res Technol, 6(9): 141-143.
[32]	Rodrigues L A, Cañizares L D C C, Meza S L R, et al. 2024. A review of the influence of genotype, environment, and food processing on the bioactive compound profile of red rice (Oryza sativa L.). Agronomy: Basel, 14(3): 616.
[33]	Santana D C, Cotrim M F, Flores M S, et al. 2021. UAV-based multispectral sensor to measure variations in corn as a function of nitrogen topdressing. Remote Sens Appl: Soci Environ, 23:100534.
[34]	Sarker I H. 2021. Machine learning: Algorithms, real-world applications and research directions. SN Comput Sci, 2(3): 160.
[35]	Štepanovský M, Ibrová A, Buk Z, et al. 2017. Novel age estimation model based on development of permanent teeth compared with classical approach and other modern data mining methods. Forensic Sci Int, 279: 72-82.
[36]	Taleon V, Gallego S, Orozco J C, et al. 2020. Retention of Zn, Fe and phytic acid in parboiled biofortified and non-biofortified rice. Food Chem X, 8: 100105.
[37]	Wang L S, Wang W S, Huang Z L, et al. 2024. Discrimination of internal crack for rice seeds using near infrared spectroscopy. Spectrochim Acta Part A: Mol Biomol Spectrosc, 319: 124578.
[38]	Wu K, Zhang Z L, He X H, et al. 2025. Using visible and NIR hyperspectral imaging and machine learning for nondestructive detection of nutrient contents in sorghum. Sci Rep, 15(1): 6067.
[39]	Zhang L, Sun H, Li H, et al. 2021. Identification of rice-weevil (Sitophilus oryzae L.) damaged wheat kernels using multi-angle NIR hyperspectral data. J Cereal Sci, 101: 103313.